Economics of radiology AI: Strategies to get paid
As artificial intelligence (AI) rapidly spreads across radiology departments, one of the biggest unanswered questions remains how healthcare providers and vendors will actually get paid for using the technology.
Speaking at the annual meeting of the Radiological Society of North America (RSNA), Nina Kottler, MD, MS, FSIIM, FAIM, chief medical AI officer for Mosaic Clinical Technologies, said reimbursement pathways for radiology AI remain fragmented and difficult to navigate because U.S. healthcare payment systems are split across multiple care settings.
“We've got a payment for inpatients. We've got a separate payment system for outpatients. We have a third payment system for non-hospital outpatients and physicians,” Kottler explained. “To find something that overlies all of them is very difficult.”
Kottler said one of the biggest barriers to physician reimbursement is the structure of the Medicare Physician Fee Schedule, which is budget neutral. Adding reimbursement for new AI technologies means reducing payments elsewhere, creating competition among specialties for limited funds.
As a result, most successful AI reimbursement efforts have focused instead on hospital payment systems.
For inpatient care, vendors increasingly are seeking reimbursement through the New Technology Add-on Payment (NTAP) program. This allows hospitals to receive temporary additional payments for qualifying technologies that improve patient care or offset costs not fully covered by existing diagnosis-related group (DRG) payments.
Kottler said the pathway has become more attractive as more AI developers obtain breakthrough device designation from the U.S. Food and Drug Administration, which can help support NTAP eligibility.
“It basically can help increase your DRG if you're losing money on that,” she said.
For outpatient imaging, AI vendors are increasingly pursuing New Technology Ambulatory Payment Classification (APC) add-on payments. These temporary reimbursement mechanisms apply to hospital-owned outpatient imaging centers and may not require a permanent Category I CPT billing code.
However, Kottler said obtaining long-term physician reimbursement through CPT coding remains challenging because many AI applications simply automate findings radiologists are already expected to identify.
“If your model is detecting something on a chest CT, well, that's already being paid for,” she said. “You're not going to get a Category I CPT code.”
The imbalance between hospital and physician reimbursement also highlights broader tensions in AI economics. Kottler noted that hospitals may receive hundreds of dollars in additional reimbursement tied to AI-enabled imaging exams, while radiologists interpreting those studies often receive only modest professional fees.
“The radiologist has to interpret that study with the AI,” she said. “How much does the radiologist get paid? On the order of around $30 to $34.”
ROI can be another way to justify costs of AI
Because direct reimbursement remains limited, many healthcare systems instead justify AI adoption through return-on-investment calculations. According to Kottler, earlier generations of radiology AI primarily generated value through improved disease detection that led to additional downstream procedures and follow-up imaging. Catching more disease early can also help reduce a practice's liability profile for a missed diagnosis.
More recently, however, the focus has shifted toward workflow efficiency and capacity generation as imaging volumes continue to outpace workforce growth.
“It’s about a 4% growth in imaging per year and capacity growth is maybe 0.4, one-tenth of that,” Kottler said. “We have to find something that not just brings more imaging into the workflow, but actually helps solve making the capacity more manageable.”
She emphasized that radiologists are not looking for tools that simply increase workloads, but rather systems that improve efficiency while reducing fatigue.
“We're looking for tools that make them more efficient,” she said. “Now efficient means you can read more studies, but it's less work.”
Kottler said the industry is beginning to transition away from narrow AI tools focused on single detection tasks, such as identifying pneumothorax or pulmonary nodules, toward more comprehensive “foundation models” and vision-language systems capable of assisting across entire radiology workflows.
These newer systems can analyze imaging exams, review prior reports and draft radiology findings in a way that resembles a radiology resident assisting with interpretation.
“The capacity generation that we're seeing on these more holistic systems, that's a game changer,” Kottler said.
She also pointed to growing interest in opportunistic imaging AI applications, such as automatically identifying coronary calcium on chest CT scans performed for other reasons. While such findings may not generate direct reimbursement, they can create downstream clinical value for both fee-for-service and value-based health systems by identifying disease earlier and keeping patients within integrated care networks.
“We need to move away from diagnosis when it's too late to detection early on,” Kottler said.
In addition to hospital reimbursement and provider return on investment, Kottler said a third emerging payment pathway involves direct patient-pay AI services. Some breast imaging practices, for example, are offering AI-enhanced cardiovascular risk assessments based on breast arterial calcifications identified during mammography exams.
However, Kottler cautioned that patient-pay models should focus on delivering genuinely new clinical information rather than charging patients for services already included in standard imaging interpretation.
As AI adoption accelerates across radiology, Kottler said future success will depend less on isolated algorithms and more on comprehensive systems capable of improving efficiency, reducing burnout and helping healthcare systems manage surging imaging demand.
Dave Fornell has covered healthcare for more than 17 years, with a focus in cardiology and radiology. Fornell is a 5-time winner of a Jesse H. Neal Award, the most prestigious editorial honors in the field of specialized journalism. The wins included best technical content, best use of social media and best COVID-19 coverage. Fornell was also a three-time Neal finalist for best range of work by a single author. He produces more than 100 editorial videos each year, most of them interviews with key opinion leaders in medicine. He also writes technical articles, covers key trends, conducts video hospital site visits, and is very involved with social media. E-mail: [email protected]